I am working on a project in a small team to build a viewer for large ships. These ships have between 50000 and 150000 components. We merged the geometries into 250 meshes to reduce drawcalls.
Right now we have a RAM usage of approximately 10GB. What are some techniques we can use to reduce RAM usage? At the moment we are implementing instance meshing, however the improvement in memory usage seem very small for our models
Merging geometries not only increases the memory usage, but also disables frustum culling. Make sure you’re merging only geometries that make sense to be merged, not everything all at once.
This also sounds like a classical usecase for instancing/InstancedMesh.
Hi @Geert iam also working in a project like yours and i cant even load a 2gb fbx file smoothly
can i look at your code ?
It’s not viable to load a 2GB file in the browser - even if you manage to do that locally, users would need to download that entire file for the parsing to even start.
Consider splitting and optimising the file.